Date: Tue, 29 Jul 1997 22:10:36 +0200 From: Stefan Esser <se@FreeBSD.ORG> To: =?iso-8859-1?Q?S=F8ren_Schmidt?= <sos@sos.freebsd.dk> Cc: freebsd-current@FreeBSD.ORG Subject: Re: code talks: announcing EIDE bus master patches Message-ID: <19970729221036.52544@mi.uni-koeln.de> In-Reply-To: =?iso-8859-1?Q?=3C199707291949=2EVAA00271=40sos=2Efreebsd=2Edk=3E=3B_fro?= =?iso-8859-1?Q?m_S=F8ren_Schmidt_on_Tue=2C_Jul_29=2C_1997_at_09=3A49=3A2?= =?iso-8859-1?Q?9PM_%2B0200?= References: <19970729210723.18104@mi.uni-koeln.de> <199707291949.VAA00271@sos.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
On Jul 29, Søren Schmidt <sos@sos.freebsd.dk> wrote: > > I'd love to see Bonnie results for them > > with the bus-master EIDE driver ... > > > > You don't have, by chance, any numbers > > to share ? :) > > Sure, on my dev machine (P6@233Mhz-Natoma/64MB/2*Maxtor 84000A6) This seems to be an OEM version of the 83840a6, if I interpret the data sheets on their web page correctly. AFAIK the fastest EIDE drive series on the market today! > First without busmatering DMA: > > -------Sequential Output-------- ---Sequential Input-- --Random-- > -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- > Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU > 100 4871 35.9 4738 9.4 1722 3.6 4488 37.2 6973 8.5 126.7 2.2 Well, as was discussed before, the PIO transfers take place mostly in an interrupt handler, which prevents accounting of the cycles to the process that caused them ... > Then with busmastering DMA: > > -------Sequential Output-------- ---Sequential Input-- --Random-- > -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- > Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU > 100 5225 39.8 5278 11.1 2017 4.4 7315 60.2 8737 10.8 149.6 2.2 And the result is apparently *higher* CPU load, at least if you (did) trust the numbers reported by Bonnie ... But the "per char" read performance improvement indicates what's really going on: CPU cycles spent in the interrupt handler probably account for the 4.5MB/s mark in the PIO case, while raw data rate of the disk is approached with the DMA driver. But writes are still significantly slower than reads, which makes me think, that one revolution of the media is lost per 64KB written ... > Not bad, for a "simple" software upgrade :) True! But there is still room for improvement ... ;) > It has improved my worldstone by ~10% if that counts for real world use.. Yes, one thing that is often underestimated is the fact, that while PIO mode transfers consume only a small fraction of the cycles of a fast CPU, but they tend to consume them exactly at the time, when the CPU is highly loaded anyway. Thanks for sending those very interesting Bonnie results! Regards, STefan
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19970729221036.52544>